An unmanned helicopter autonomous return decision system

Abstract

The application relates to the technical field of unmanned aerial vehicle autonomous navigation, and discloses an unmanned helicopter autonomous return decision system. After triggering autonomous return, the autonomous return decision system takes a return triggering anchor point as a backtracking terminal to obtain navigation evidence records, performs evidence attribution encapsulation and time connection encapsulation, and generates a historical navigation evidence sequence to be reviewed. Then, the autonomous return decision system constructs a return candidate track segment according to changes in a tower-circulating inspection action stage and flight mode switching before return triggering. Subsequently, the autonomous return decision system performs closed review based on target wind turbine structure observation changes and track relative motion changes, and identifies a locally self-consistent pollution segment. Finally, the autonomous return decision system forms a return connection anchor point, a pollution boundary segment and a prohibited reverse use segment according to the local self-consistent pollution identification result, and writes the return path generation control record. The autonomous return decision system can avoid that a historical track with continuous values but an open structure is mistaken as return starting basis, and improves the reliability and safety of unmanned helicopter autonomous return path generation.

Description

Technical Field

[0001] This application relates to the field of autonomous navigation technology for unmanned aerial vehicles (UAVs), and in particular to an autonomous return-to-home decision-making system for unmanned helicopters. Background Technology

[0002] During offshore wind farm inspections, unmanned helicopters typically take off from maintenance vessels or offshore platforms and perform low-altitude fly-around inspections around the target wind turbine. Taking blade surface inspection as an example, when the unmanned helicopter approaches the outer edge of the tower and blade sweep area, it needs to keep the camera payload continuously pointed at the leading edge of the blade, the blade root connection area, or the outer wall of the tower. The helicopter's flight path may move along the circumference of the tower, but the nose direction remains pointed towards the target being photographed for a long time. This flight attitude differs significantly from ordinary straight-line cruising. When the return flight is triggered, the unmanned helicopter is often still in a state of coupled lateral fly-around, attitude adjustment, and payload alignment. At the same time, sea surface reflection... This will change the propagation path of satellite navigation signals. The metal structure of the wind turbine tower and nacelle will cause obstruction and multipath effects. The change in the position of the maintenance vessel will also give the return target dynamic attributes. Existing unmanned helicopter return systems usually form a return path based on the position results output by the airborne navigation module, historical tracks, and return point information. When the navigation environment is disturbed, the return capability is maintained through methods such as combined navigation, track smoothing, abnormal position elimination, and historical path backtracking. For offshore operation scenarios, some solutions will also combine mobile platform positioning and visual sign recognition to update the return destination, so as to improve the feasibility of returning to the vicinity of the maintenance vessel.

[0003] In offshore wind turbine tower inspection scenarios, the positioning result at the moment of triggering the return-to-base operation cannot necessarily be directly used as the starting point for the return-to-base calculation. For example, when an unmanned helicopter is photographing the blades on the leeward side of the tower, the communication link quality may decrease due to tower obstruction, and the battery level may also enter the return-to-base protection zone at the same time, thus triggering the system to autonomously return to base. At this time, the flight path before the trigger may still be numerically continuous, and the changes in flight speed, altitude, and attitude may also conform to the characteristics of tower inspection. However, the overall spatial position has already been offset due to the influence of multipath on the sea surface and the obstruction of the wind turbine structure. If the system starts the return-to-base operation based solely on the positioning result at the moment of triggering, coordinates with the risk of offset will continue to be included in the calculation of the return-to-base direction, and the remaining coordinates will be affected. The process of range assessment and return path generation makes subsequent decisions based on unstable navigation anchor points. When relying solely on jump point elimination and filtering smoothing, the ability to identify such locally continuous but overall offset track states is limited. When relying solely on historical track backtracking, it is also difficult to distinguish which tower-circling segments can be reversed into the return decision, which segments can only be used as pollution boundary references, and which segments need to be excluded from the return path generation process. Therefore, the autonomous return decision of offshore wind power inspection unmanned helicopters still needs to determine the eligibility of historical navigation evidence at the return trigger stage to prevent pollution trigger points and unreversible tower-circling tracks from continuing to be transmitted to the return direction, remaining range, and return path judgment. Summary of the Invention

[0004] This application proposes an autonomous return-to-home decision system for unmanned helicopters to address the problems mentioned in the background art.

[0005] To achieve the above objectives, this application adopts the following technical solution: an autonomous return decision system for unmanned helicopters, comprising: a navigation evidence encapsulation module, which acquires navigation evidence records formed by an unmanned helicopter for offshore wind power inspection before the return trigger anchor point, wherein the return trigger anchor point corresponds to the autonomous return trigger time, and generates a sequence of historical navigation evidence to be reviewed by adopting an evidence encapsulation method with the return trigger anchor point as the retrospective endpoint; the navigation evidence records characterize the correspondence between the stage of the inspection around the target wind turbine and the navigation measurement results within the same retrospective time period; The candidate segment construction module uses the tower-around action boundary segmentation method to process the historical navigation evidence sequence to be reviewed and generate return candidate track segments. The tower-around action boundary is formed by the changes in the stage of the tower-around inspection action around the target wind turbine and the flight mode switching before the return is triggered. The geometric closure verification module adopts a closure verification method consisting of the relative motion changes of the trajectory derived from the observed changes in the target wind turbine structure and the candidate trajectory segments of the return flight. It processes the candidate trajectory segments of the return flight and generates local self-consistent pollution identification results. The return-to-base permit output module performs reverse use qualification determination on the candidate return-to-base track segments based on the local self-consistent pollution identification results, generates the return-to-base start-up permit result, and writes the candidate return-to-base track segments into the return-to-base path generation control record based on the return-to-base start-up permit result. The return-to-base path generation control record includes the return-to-base acceptance anchor point, pollution boundary segments, and prohibited reverse use segments.

[0006] Furthermore, the specific operation of the evidence encapsulation method is as follows: read the airborne return trigger record used to form the return trigger anchor point, and take the record time of forming the autonomous return command as the return trigger anchor point; take the return trigger anchor point as the backtracking endpoint, and generate the backtracking period based on the base duration of the tower-circling inspection action in the current inspection task configuration and the readable record coverage duration before the return trigger; During the retrospective period, candidate records belonging to the same target wind turbine surrounding tower inspection action phase are extracted. Based on the task attribution of the candidate record relative to the return trigger anchor point and the execution evidence attribution of the action phase, the candidate records with the established attribution relationship are marked as action phase locked records, and the candidate records with the invalid attribution relationship are marked as non-locked evidence records.

[0007] Furthermore, the generation operation of the historical navigation evidence sequence to be reviewed is as follows: The execution time of the action phase lock record is encapsulated, and the action observation time is matched with the navigation measurement time in the navigation measurement record within the allowable time deviation. The allowable time deviation is formed based on the airborne recording time synchronization error and the sampling period of the navigation measurement record. When all navigation measurement times adjacent to the action observation time exceed the allowable time deviation, a low-frequency measurement marker is written, and the corresponding candidate record is encapsulated as an unencapsulated evidence unit. When a match is established, an evidence acceptance value is generated according to the time deviation level between the action observation time and the navigation measurement time, and the matched candidate record is encapsulated as a historical navigation evidence unit. The navigation evidence encapsulation module arranges the historical navigation evidence units and unencapsulated evidence units according to the chronological order of the action observation times, generating the historical navigation evidence sequence to be reviewed.

[0008] Furthermore, the boundary candidate generation operation of the tower-around action boundary segmentation method is as follows: scan the historical navigation evidence sequence to be reviewed in chronological order of action observation time. The evidence units in the historical navigation evidence sequence to be reviewed include historical navigation evidence units and unaccepted evidence units. Based on the action observation time of the evidence unit, corresponding candidate records and flight control mode records before the return trigger anchor point are obtained within the same online recording period. The corresponding candidate records form the action phase type, and the flight control mode records form the flight mode field. The action phase transition and flight mode switch are used as the trigger conditions for the tower-around action boundary. The action phase transition is jointly confirmed by the action phase type of the adjacent evidence unit and the action phase continuous confirmation condition, and the flight mode switch is confirmed by the flight mode field before the return trigger anchor point. When the tower-around action boundary trigger condition is formed, the action observation time corresponding to the trigger condition is written into the tower-around action boundary candidate record. The action phase continuous confirmation condition is formed by the action record sampling period and the shortest action hold time in the current inspection task configuration.

[0009] Furthermore, the generation operation of the return candidate track segment is as follows: based on the time continuity status of the evidence units in the historical navigation evidence sequence to be reviewed, boundary restriction processing is performed on the candidate records of the tower-circling action boundary to form internal boundary records. The time continuity status is jointly characterized by the evidence continuity value of the historical navigation evidence unit and the non-continuation status of the non-continuation evidence unit. The internal boundary records within the same action record sampling period are merged and, together with the sequence start boundary and sequence end boundary of the historical navigation evidence sequence to be reviewed, form a segment encapsulation boundary sequence. The sequence start boundary and sequence end boundary are formed by the action observation time of the first evidence unit and the last evidence unit in the historical navigation evidence sequence to be reviewed, respectively. Fragment encapsulation is performed on the evidence unit set between adjacent fragment encapsulation boundaries. When the evidence unit set contains only unaccepted evidence units, a fragment insufficient acceptance record is written. When the evidence unit set contains historical navigation evidence units, the fragment is encapsulated to generate a return candidate track fragment. Unaccepted evidence units are not considered as valid continuous evidence of relative motion changes within the track.

[0010] Furthermore, the operations for forming the observed changes in the target wind turbine structure and the relative motion changes in the trajectory are as follows: The segment start boundary, segment end boundary, navigation measurement result reference marker, and segment status record of the return candidate trajectory segment are read, and the navigation measurement results corresponding to the historical navigation evidence units within the segment are called according to the navigation measurement result reference marker; the prior record of the target wind turbine structure and the load observation record in the same online recording period as the return candidate trajectory segment are read, and the projection relationship of the target wind turbine stable structure object corresponding to the segment start boundary and segment end boundary is compared before and after to form the observed changes in the target wind turbine structure; using the target wind turbine reference position in the prior record of the target wind turbine structure as the relative reference object, the relative position conversion of the navigation measurement results is performed to form the relative motion changes in the trajectory.

[0011] Furthermore, the specific operation of the closed-loop verification method is as follows: based on the segment status record, the correspondence between the target wind turbine stable structure object and the mission navigation coordinate reference, a closed-loop verification admission judgment is performed on the return candidate track segment; when the closed-loop verification admission judgment fails, the return candidate track segment is marked as a closed-loop verification restricted segment; when the closed-loop verification admission judgment passes, directional closed-loop comparison and radial closed-loop comparison are performed on the observed changes in the target wind turbine structure and the relative motion changes of the track to form the closed-loop deviation.

[0012] Furthermore, the operation for generating the local self-consistent pollution identification results is as follows: read the closure deviation amount and the closure deviation judgment boundary. The closure deviation judgment boundary is formed by the stable tower inspection records of the same type of unmanned helicopter in the same type of target wind turbine inspection mission. When historical samples are insufficient, the initial judgment boundary in the current inspection mission configuration is used. When the closure deviation is not greater than the closure deviation judgment boundary and the return candidate track segment is not marked as a closure verification restricted segment, the return candidate track segment is marked as a closure consistent segment; when the closure deviation is greater than the closure deviation judgment boundary, the navigation measurement continuity is confirmed by the time continuity status and the navigation measurement result reference mark, and the return candidate track segment is not marked as a closure verification restricted segment, the return candidate track segment is marked as a locally self-consistent contaminated segment, and the segment start boundary and segment end boundary of the locally self-consistent contaminated segment are written into the contaminated boundary candidate result.

[0013] Furthermore, the specific operation for reverse use qualification determination is as follows: using the segment start boundary and segment end boundary of the return candidate track segment as indexes, a correspondence is established between the local self-consistent contamination identification result and the return candidate track segment; when the same return candidate track segment is simultaneously marked as two or more of the following: local self-consistent contamination segment, closed verification restricted segment, and closed consistency segment, the valid segment status is determined according to the priority order of local self-consistent contamination segment, closed verification restricted segment, and closed consistency segment; the return candidate track segment whose segment end boundary is no later than the return trigger anchor point and whose time distance from the return trigger anchor point is closest is determined as the reverse scan start segment, and a reverse scan is performed from the reverse scan start segment towards the historical time direction; the scan stops when the first local self-consistent contamination segment or closed verification restricted segment appears, and the closed consistency segment before the stop is used to generate the return acceptance anchor point.

[0014] Furthermore, the generation operations for the return-to-origin permit calculation result and the return-to-origin path generation control record are as follows: In the closed and consistent segment before stopping, the target closed and consistent segment is determined according to the segment boundary sequence relationship, and the return-to-origin anchor point is formed by the segment end boundary of the target closed and consistent segment and the navigation measurement result reference mark corresponding to the historical navigation evidence unit with valid time continuity adjacent to the segment end boundary; a contaminated boundary segment is formed based on the contaminated boundary candidate result and the segment boundary of the locally self-consistent contaminated segment; a prohibited reverse-use segment is formed based on the return-to-origin candidate track segment corresponding to the stop of the reverse scan, and the return-to-origin candidate track segment that is interrupted by any segment of the contaminated boundary segment or the closed verification restricted segment of the reverse call path; the return-to-origin permit output module generates the return-to-origin permit calculation result based on the return-to-origin anchor point, the contaminated boundary segment, and the prohibited reverse-use segment, and writes the return-to-origin anchor point, the contaminated boundary segment, and the prohibited reverse-use segment into the return-to-origin path generation control record.

[0015] The beneficial effects of this invention are as follows: This invention replaces the historical flight path of an unmanned helicopter for offshore wind power inspection before its return flight with a directly retrievable continuous navigation record. Instead, it first creates verifiable candidate flight path segments by dividing the data based on evidence attribution, time continuity, and tower-circling action boundaries. Then, it uses observations of changes in the stable structure of the target wind turbine to perform a closed-loop verification of the relative motion changes of the flight path. This identifies local self-consistent contamination segments that are continuous within the navigation values ​​but not closed relative to the target wind turbine structure. In scenarios such as tower obstruction, nacelle metal reflection, sea surface reflection, or link protection, this invention avoids the system selecting an incorrect return flight starting point based solely on the continuity of navigation measurement results. Furthermore, it controls the return flight path generation process by using return flight anchor points, contamination boundary segments, and prohibited reverse-use segments, making the calculation basis for the unmanned helicopter's autonomous return flight more reliable and improving the return flight safety and path generation stability in offshore wind power inspection missions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort: Figure 1 This is a system framework diagram of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] like Figure 1 As shown, the present invention discloses an autonomous return-to-home decision system for unmanned helicopters, comprising: a navigation evidence encapsulation module, a candidate fragment construction module, a geometric closure verification module, and a return-to-home permission output module.

[0019] The navigation evidence encapsulation module is used to acquire navigation evidence records formed before the return trigger anchor point after the unmanned helicopter for offshore wind power inspection triggers autonomous return. It adopts an evidence encapsulation method with the return trigger anchor point as the traceback endpoint to generate a sequence of historical navigation evidence to be reviewed. The navigation evidence records represent the correspondence between the inspection action phase around the target wind turbine and the navigation measurement results in the same traceback period. The navigation evidence encapsulation module organizes the online records formed before the return trigger into evidence objects that can be segmented and reviewed later. The generated sequence of historical navigation evidence to be reviewed is used as the input of the candidate fragment construction module.

[0020] In offshore wind power inspection scenarios, unmanned helicopters typically perform external inspections of the blades, towers, or nacelles of the target wind turbine. During blade leading edge inspection, when the unmanned helicopter maintains circumferential flight around the outside of the target wind turbine tower, the camera payload continuously points towards the blade leading edge or blade root area. The aircraft's flight path moves along the periphery of the target wind turbine, and the nose direction may deviate from the actual flight path direction. When the unmanned helicopter moves to the leeward side of the tower or near the nacelle's obstructed area, the link quality degrades, the battery enters the protection zone, or the wind speed triggers safety protection. The flight control system generates an autonomous return command. At this time, continuous navigation measurement records may still be retained near the return trigger anchor point. These records are first encapsulated with attribution and time inheritance to ensure that the boundary segmentation of subsequent tower circling actions is established under the same time reference and the same target wind turbine object.

[0021] First, the airborne return-to-home trigger record used to form the return-to-home trigger anchor point is read. The airborne return-to-home trigger record is formed by the unmanned helicopter flight control system when it switches from inspection mode to autonomous return-to-home mode, and it serves as the endpoint for the backtracking of this step. The navigation evidence encapsulation module uses the record time of forming the autonomous return-to-home command as the return-to-home trigger anchor point. If the trigger records of power protection, link quality degradation and wind speed safety protection occur successively within the same return-to-home cycle, the navigation evidence encapsulation module uses the record time of actually writing the autonomous return-to-home command as the unique return-to-home trigger anchor point, so that subsequent historical record readings revolve around the same time reference.

[0022] A backtracking period is generated with the return-to-base trigger point as the backtracking endpoint. The backtracking period is jointly limited by the baseline duration of the tower-around inspection action in the current inspection task configuration and the readable record coverage duration before the return-to-base trigger. The baseline duration of the tower-around inspection action serves as the target read length, while the readable record coverage duration before the return-to-base trigger serves as the upper limit of readable records. Preferably, the backtracking period length is 10 to 90 seconds, measured in seconds. Its value is jointly limited by the baseline duration of the tower-around inspection action in the current inspection task configuration, the readable record coverage duration of the onboard cache before the return-to-base trigger, and the stable tower-around inspection records of similar wind turbines. The baseline duration of the tower-around inspection action ensures that the backtracking period can cover the stable tower-around inspection action before the return-to-base trigger and its adjacent state transition records, while the readable record coverage duration limits the upper limit of the actual historical records that can be read. When performing the same type of inspection task for the first time and the current inspection task configuration does not record the reference duration of the tower-circling inspection action, the navigation evidence encapsulation module uses the inspection task template duration formed by the same type of unmanned helicopter for the same type of target wind turbine, or uses the inspection script duration confirmed by the task configuration as the initial reference duration of the tower-circling inspection action. After the subsequent actual inspection records reach the number of stable tower-circling inspections required by the task configuration, the navigation evidence encapsulation module updates the reference duration based on the actual stable tower-circling inspection records. The updated reference duration of the tower-circling inspection action is used for subsequent return cycles and does not write back the historical navigation evidence sequence to be reviewed that has been generated in the current return cycle. If the readable record coverage duration before the return is triggered is shorter than the reference duration of the tower-circling inspection action, the navigation evidence encapsulation module reads the historical records according to the readable record coverage duration and writes the low evidence coverage mark.

[0023] During the retrospective period, candidate action records related to the inspection task before the return-to-base trigger anchor point are extracted. These candidate action records are derived from online action records already formed before the return-to-base trigger anchor point, and must have at least an action observation time. The action observation time is derived from the target wind turbine's surrounding tower inspection action phase record and serves as the time reference for the action phase. The navigation measurement record records the navigation measurement results and their navigation measurement time, and is derived from the airborne navigation record that has completed online time synchronization, serving as the reference for the navigation measurement results. During the time-series encapsulation phase, the action observation time is matched with the navigation measurement time in the navigation measurement record. When the target wind turbine number, inspection task number, and tower inspection action phase record all correspond to the return-to-base trigger anchor point, the candidate action record is encapsulated as evidence attribution. Records located within the retrospective period but corresponding to the previous wind turbine inspection action or the action that was disengaged after the return-to-base trigger are marked as non-locked evidence records.

[0024] Based on the task affiliation and action phase affiliation of the action-side candidate record relative to the return trigger anchor point, the navigation evidence encapsulation module marks the action-side candidate record as an action phase locked record if the action-side candidate record and the return trigger anchor point belong to the same inspection task and belong to the same target wind turbine surrounding tower inspection action phase. If the action-side candidate record does not simultaneously satisfy the task affiliation and action phase affiliation, the navigation evidence encapsulation module marks the action-side candidate record as an unlocked evidence record. The action phase locked record enters the subsequent time inheritance encapsulation, while the unlocked evidence record is not used as a valid inheritance object and enters the pending historical navigation evidence sequence. Its affiliation is not established in the evidence encapsulation log. The action phase locked mark is used to indicate whether the candidate record belongs to the target wind turbine surrounding tower inspection action phase before this return trigger and serves as the admission mark for the candidate fragment construction module to identify valid action evidence.

[0025] For the action phase lock record, the navigation evidence encapsulation module performs time-based encapsulation. Time-based encapsulation takes the time correspondence between the action observation time and the navigation measurement time as the processing object. The navigation evidence encapsulation module uses the online records that were formed before the return trigger anchor point and excludes the verification results after the return trigger from the encapsulation judgment in this step. In the offshore wind power tower inspection, when the unmanned helicopter flies laterally around the outside of the tower, the azimuth of the aircraft relative to the target wind turbine changes rapidly in a short period of time. The action observation time and the navigation measurement time are within the same time scale, so that the subsequent candidate fragment construction module can call the action change and the navigation measurement result as the same evidence unit.

[0026] The action observation time is matched with the navigation measurement time in the navigation measurement record within the allowable acceptance time deviation. The allowable acceptance time deviation is formed based on the airborne recording time synchronization error and the sampling period of the navigation measurement record. The airborne recording time synchronization error comes from the online time synchronization calibration record between the flight control record, the inspection load record, and the navigation measurement record. The sampling period of the navigation measurement record comes from the actual sampling period of the airborne navigation record. The allowable acceptance time deviation, the action record sampling period used by the time break marker, and the low acceptance level used by the acceptance break marker are all based on the same set of online time synchronization calibration records and sampling period records.

[0027] In a typical online reception scenario, the allowable reception time deviation is preferably between 0.1 and 0.5 seconds. When the sampling period of the navigation measurement record is long, and the inspection speed range and flight control speed record in the current inspection task configuration jointly indicate that the unmanned helicopter is within the stable tower circling speed range, the allowable reception time deviation can be extended to no more than 1 second. The tower circling displacement corresponding to the extended time deviation must be within the resolvable range of the action record. The resolvable range of the action record is formed by the action record sampling period and the stable tower circling speed range. If the navigation measurement times adjacent to the action observation time all exceed the allowable reception time deviation, or the tower circling displacement corresponding to the extended time deviation exceeds the resolvable range of the action record, the navigation evidence encapsulation module writes low-frequency measurement markers and non-reception markers, and encapsulates the corresponding candidate records as non-reception evidence units. The candidate records written with low-frequency measurement markers enter the historical navigation evidence sequence to be reviewed as non-reception evidence units, serving as boundary restriction information for the candidate fragment construction module, and do not participate in the formation of continuous boundaries as historical navigation evidence units.

[0028] When the action observation time matches the navigation measurement time, the navigation evidence encapsulation module generates an evidence acceptance value according to the time deviation level between the action observation time and the navigation measurement time. The time deviation level is based on the allowable acceptance time deviation. When the time deviation does not exceed 1 / 4 of the allowable acceptance time deviation, the evidence acceptance value is 1; when the time deviation is greater than 1 / 4 of the allowable acceptance time deviation but not more than 1 / 2 of the allowable acceptance time deviation, the evidence acceptance value is 0.8; when the time deviation is greater than 1 / 2 of the allowable acceptance time deviation but not more than the allowable acceptance time deviation, the evidence acceptance value is 0.5; when the time deviation exceeds the allowable acceptance time deviation, an unaccepted state is formed, and the evidence acceptance value is 0. The evidence acceptance value is used to indicate the degree of acceptance between the target wind turbine surrounding tower inspection action stage and the navigation measurement results on the time scale. The value ranges from 0 to 1 and plays a role in boundary continuity constraint in the candidate segment construction module.

[0029] After completing the time-based encapsulation, the matching candidate records on the action side are encapsulated into historical navigation evidence units. The historical navigation evidence units contain the action observation time, navigation measurement time, action phase lock mark, evidence acceptance value, and navigation measurement result reference mark. The navigation measurement result reference mark is used to point to the original navigation measurement record, so that the subsequent candidate segment construction module and geometric closure verification module can trace the navigation measurement results along the same record source. When forming historical navigation evidence units, the navigation evidence encapsulation module retains the reference relationship of the original navigation measurement results, so that the navigation measurement results with continuous values ​​in the leeward side of the tower triggering return condition are still in the state of pending verification.

[0030] For candidate action records that fail to complete time continuity matching, the navigation evidence encapsulation module encapsulates the candidate action record as an unconnected evidence unit. The unconnected evidence unit retains the action observation time, low-frequency measurement marker, and unconnected marker, which indicates that although the record is located within the retrospective period and belongs to the target wind turbine surrounding tower inspection action stage, its action observation and navigation measurement failed to establish a time continuity relationship within the allowable connection time deviation. The unconnected evidence unit is retained in the sequence as a constraint information of the tower-circling action boundary formed by the candidate segment construction module, and is used to restrict subsequent steps from interpreting the record gap as continuous tower-circling action.

[0031] Evidence units are arranged in chronological order of action observation times to generate a sequence of historical navigation evidence to be reviewed. Evidence units include historical navigation evidence units and unaccepted evidence units. The time range of the historical navigation evidence sequence to be reviewed is within the backtracking period with the return trigger anchor point as the backtracking endpoint. If the interval between action observation times of adjacent evidence units exceeds twice the action record sampling period, the navigation evidence encapsulation module writes a time break marker. Twice the action record sampling period comes from the action record sampling period in the same online time synchronization calibration record and is used to accommodate the absence of a single action sampling point. It is not used to relax the acceptance matching between action observation and navigation measurement. If more than two consecutive evidence units are in a low acceptance level or unaccepted state, the navigation evidence encapsulation module writes an acceptance break marker. Both the time break marker and the acceptance break marker serve as candidate segment construction modules to form the restriction information of the around-tower action boundary. The time break marker is used to restrict continuous segmentation across break records, and the acceptance break marker is used to restrict low acceptance records and unaccepted records from continuously entering the same around-tower action boundary.

[0032] Regarding anomaly handling, if the return-to-base trigger anchor point is missing, a missing return-to-base trigger anchor point marker is output, and the evidence encapsulation process for this branch is stopped. If the number of valid historical navigation evidence units in the historical navigation evidence sequence to be reviewed is lower than the minimum number of evidence units required by the candidate segment construction module to form a tower-around action boundary, the navigation evidence encapsulation module generates a low evidence coverage marker and retains the evidence units that have been encapsulated. The minimum number of evidence units is formed by the reference duration of the tower-around inspection action and the sampling period of the action record. If the action observation time and the navigation measurement time do not complete the connection matching, the corresponding action-side candidate record is encapsulated as an unconnected evidence unit and written with an unconnected marker. The above-mentioned anomaly markers serve as input restrictions when the candidate segment construction module performs tower-around action boundary segmentation, so that insufficient historical records, low-frequency measurements, and time breaks all enter subsequent processing in the form of evidence status.

[0033] The historical navigation evidence sequence to be reviewed is used to provide the candidate segment construction module with the evidence continuation basis within the same retrospective period. The candidate segment construction module determines the candidate position of the tower-circling action boundary based on the action phase lock mark, non-acceptance mark, low-frequency measurement mark, time break mark and acceptance break mark in the historical navigation evidence sequence to be reviewed. The navigation evidence encapsulation module completes the attribution encapsulation and time continuation encapsulation of the historical records, forming a historical navigation evidence sequence to be reviewed with task attribution, action phase attribution and time continuation status, so that the subsequent tower-circling action boundary segmentation can be carried out based on the evidence continuation relationship within the same retrospective period.

[0034] The candidate fragment construction module receives the historical navigation evidence sequence to be reviewed from the navigation evidence encapsulation module and processes it using the tower-around action boundary segmentation method to generate return candidate track fragments. The tower-around action boundary is formed by the changes in the phase of the tower-around inspection action around the target wind turbine and the flight mode switching before the return is triggered. The processing object of the candidate fragment construction module is the evidence unit sequence that has completed the attribution encapsulation and time succession encapsulation. The evidence unit includes historical navigation evidence units and unsuccessful evidence units. The return candidate track fragment enters the geometric closure review module as the fragment object for performing closure review of the relative motion change of the track and the observed changes in the target wind turbine structure.

[0035] During offshore wind turbine tower inspections, when unmanned helicopters perform actions such as blade positioning, tower circumferential flight, load alignment and maintenance, return direction search, and departure from the wind turbine's near field, the aircraft's flight path, nose orientation, and load line of sight may be in different motion relationships. Before the return is triggered, the flight control system may also enter a return preparation, low-power speed limit, or link protection state due to power protection, link quality degradation, or wind speed safety protection. Fixed time length segmentation cannot distinguish the changes in the tower inspection action phases and the flight mode switch before the return is triggered. The tower inspection action boundary is defined by the action phase transition, flight mode switch, and the time continuity of the evidence unit, so that the return candidate flight path segment has a segment boundary corresponding to the tower inspection action around the target wind turbine.

[0036] The system reads the historical navigation evidence sequence to be reviewed output by the navigation evidence encapsulation module, and calls the low evidence coverage marker, time break marker, and continuity break marker already formed by the navigation evidence encapsulation module. The reading range of the candidate fragment construction module is limited to online records formed before the return trigger anchor point. If there is no evidence unit in the historical navigation evidence sequence to be reviewed, the system outputs a fragment generation failure marker and stops the fragment encapsulation processing of this module in the current return cycle.

[0037] Scan the sequence of historical navigation evidence to be reviewed in chronological order of action observation times, and obtain the corresponding candidate records and flight control mode records before the return-to-home trigger anchor point within the same online recording period based on the action observation time of the evidence unit. The same online recording period is jointly determined by the action observation time, the action recording sampling period, and the online time synchronization calibration record, which is used to limit the time attribution of the candidate records and flight control mode records relative to the evidence unit. The corresponding candidate records include at least an inspection action field or an inspection task status field to form the action phase type; the flight control mode records are used to form the flight mode field. The action phase type is used to characterize the target wind turbine surrounding tower inspection action phase corresponding to the evidence unit, and the flight mode field is used to characterize the inspection flight status, return-to-home preparation status, low-power speed limiting status, or link protection status corresponding to the evidence unit.

[0038] The transition of action phase and the switching of flight mode are used as the triggering conditions for the tower-circling action boundary. The transition of action phase is jointly confirmed by the action phase type of the adjacent evidence unit and the continuous confirmation condition of the action phase. The switching of flight mode is confirmed by the flight mode field before the return trigger anchor point. When the action phase type of the adjacent evidence unit indicates that the unmanned helicopter has changed from fixed-point shooting of the blade to circumferential flight around the tower, from load alignment and maintenance to search in the return direction, or from circumferential flight around the tower to leaving the near field of the wind turbine, it is determined that the action phase of the tower-circling inspection around the target wind turbine has changed. Evidence units with missing action phase types do not participate in the confirmation of action phase transition and are written into the action phase missing state.

[0039] The continuous confirmation condition for the action phase is formed by the action recording sampling period and the shortest action holding duration in the current inspection task configuration. The action recording sampling period is measured in seconds, and the online time synchronization calibration record and action recording configuration used by the navigation evidence encapsulation module are adopted. The number of consecutive action phase confirmations is formed by the shortest action holding duration and the action recording sampling period, preferably no less than 2 consecutive evidence units, to exclude short-term action jumps in a single evidence unit. When the converted action phase type remains consistent among the evidence units that reach the number of consecutive action phase confirmations, the corresponding action phase conversion is confirmed to be valid. When the remaining evidence units before the return trigger anchor point are insufficient to meet the continuous confirmation condition for the action phase, the candidate segment construction module confirms the corresponding boundary if there is a flight mode switching record in the same online recording period. In the case of a lack of flight mode switching records, a transient action change marker is written.

[0040] The flight mode switch before the return-to-home trigger is confirmed based on the flight mode field. When the flight mode field indicates that the unmanned helicopter has entered the return-to-home preparation state from the inspection flight state, entered the low-power speed limit state from the normal inspection speed control state, or entered the link protection state from the stable link inspection state, the flight mode switch before the return-to-home trigger anchor point is confirmed. When the trigger condition of the tower-around action boundary is formed, the action observation time corresponding to the trigger condition is written into the tower-around action boundary candidate record.

[0041] Taking the inspection of the leading edge of a target wind turbine blade as an example, the unmanned helicopter moves circumferentially outside the tower in a load-aligned manner, with the camera load continuously pointing towards the leading edge or root area of ​​the blade. When the unmanned helicopter moves to the vicinity of the leeward side of the tower and the link quality decreases, the flight control mode record switches from the inspection flight state to the link protection state. The action observation time corresponding to the action phase transition and the action observation time corresponding to the flight mode switch are written into the candidate record of the around-the-tower action boundary, so that the action changes and flight mode changes before the return trigger enter the segment boundary formation process respectively.

[0042] Based on the temporal continuity status of evidence units in the historical navigation evidence sequence to be reviewed, boundary restriction processing is performed on candidate records of tower-circling action boundaries to form internal boundary records. The temporal continuity status is jointly characterized by the evidence continuity value of the historical navigation evidence units and the non-continuation status of the non-continuation evidence units. For historical navigation evidence units with evidence continuity values ​​at a low continuity level, the candidate segment construction module uses them as evidence of continuity restriction to participate in boundary restriction processing. For non-continuation evidence units with non-continuation markers or low-frequency measurement markers, they are written as boundary restriction information into the segment status record of adjacent segments. The temporal break marker corresponds to the situation where the time interval between action observations of adjacent evidence units exceeds the allowable range of the action record sampling period, and it is used as the basis for mandatory boundaries. The continuity break marker corresponds to continuous low continuity levels or non-continuation states, and the candidate segment construction module uses it as the basis for continuity restriction boundaries.

[0043] For cases where multiple internal boundary records fall within the same action recording sampling period, boundary merging is performed. The boundary merging process uses the action recording sampling period as the minimum time resolution basis. The action recording sampling period and the time break marker in the navigation evidence encapsulation module share the same online time synchronization calibration record. If multiple internal boundary records fall within the same action recording sampling period, the multiple internal boundary records are merged into a single boundary record, and the boundary source relationship is retained in this boundary record. When a time break marker exists, the boundary time uses the action observation time corresponding to the time break marker; when no time break marker exists, the boundary time uses the action observation time corresponding to the flight mode switch; if there is no flight mode switch record within the same action recording sampling period, the boundary time uses the action observation time corresponding to the action stage type change. When the same action transition triggers action stage change, flight mode switch, and continuation break simultaneously, the candidate fragment construction module only forms one internal boundary record.

[0044] After forming the internal boundary record, the action observation time of the first evidence unit in the historical navigation evidence sequence to be reviewed is determined as the sequence start boundary, and the action observation time of the last evidence unit in the historical navigation evidence sequence to be reviewed is determined as the sequence end boundary. The sequence start boundary, internal boundary record, and sequence end boundary are then combined in the order of action observation time to form a segment encapsulation boundary sequence. If the interval between the action observation time of the last evidence unit and the return trigger anchor point meets the formation conditions of the time break mark in the navigation evidence encapsulation module, the time break source is written into the sequence end boundary at the same time. The formation conditions adopt the action record sampling period and online time synchronization calibration record in the navigation evidence encapsulation module. The sequence start boundary and sequence end boundary jointly limit the segment encapsulation range.

[0045] Fragment encapsulation is performed on the evidence unit set between adjacent fragment encapsulation boundaries. When the evidence unit set between adjacent fragment encapsulation boundaries contains only unaccepted evidence units, a fragment insufficient acceptance record is written, and this evidence unit set does not enter the complete return candidate track fragment. When the evidence unit set between adjacent fragment encapsulation boundaries contains historical navigation evidence units but the number of historical navigation evidence units does not meet the minimum evidence unit requirement, a fragment insufficient acceptance record is written, and this evidence unit set does not enter the complete return candidate track fragment. When the evidence unit set between adjacent fragment encapsulation boundaries contains historical navigation evidence units and the number of historical navigation evidence units meets the minimum evidence unit requirement, a return candidate track fragment is encapsulated and generated.

[0046] The minimum number of evidence units required is determined by the shortest action duration and action recording sampling period in the current inspection task configuration, and must be no less than 2 historical navigation evidence units. The minimum number of evidence units required is used to limit the encapsulation of intervals with insufficient evidence length into return candidate track segments, so that the return candidate track segments entering the geometric closure verification module have the basic evidence of relative motion changes of the propulsion track. When there is a low evidence coverage marker, the low coverage state is written into the segment status record. When the return candidate track segment with the low coverage state enters the geometric closure verification module, the low coverage state is read as a restriction condition for the segment.

[0047] The return candidate track segment includes the segment start boundary, segment end boundary, historical navigation evidence units within the segment, unaccepted evidence units within the segment, boundary source relationship, and segment status record. If the segment contains historical navigation evidence units and simultaneously contains low-frequency measurement markers, unaccepted markers, or acceptance break markers, the candidate segment construction module writes the acceptance limit state into the segment status record. The unaccepted evidence units are retained in the segment object as the source of the segment status record. The relative motion change of the track within the segment is formed by the navigation measurement result referenced by the navigation measurement result reference marker. The missing state of the action phase is entered into the segment status record of the return candidate track segment.

[0048] The identification of local self-consistent contamination segments and the determination of eligibility for reverse use are handled by the subsequent geometric closure verification module and the return permission output module. The candidate segment construction module determines the action boundaries and segment objects in the historical navigation evidence sequence through the triggering conditions of the tower-around action boundary, the time continuity status, the segment encapsulation boundary sequence, and the segment status record. This enables the geometric closure verification module to deduce the relative motion changes of the navigation track based on the navigation measurement result reference markers, action observation times, and segment boundaries in the return candidate track segments, and to perform a closure verification with the observed changes in the target wind turbine structure.

[0049] In this step, the same online recording period, continuous confirmation condition for the action phase, boundary merging processing, and minimum evidence unit requirement are all based on the same online time synchronization calibration record, action recording sampling period, and current inspection task configuration. Among them, the same online recording period is used to limit the time assignment of candidate records and flight control mode records relative to evidence units; the continuous confirmation condition for the action phase is used to confirm the action phase transition; the boundary merging processing is used to limit repeated segmentation within the same sampling period; the minimum evidence unit requirement is used to limit the entry of intervals with insufficient evidence length into the return candidate track segment; when historical samples are insufficient to form the shortest action holding duration, the action holding configuration in the current inspection task script is used as the initial basis and updated after subsequent stable tower-circling inspection records are accumulated; the updated shortest action holding duration is used for subsequent return cycles and does not write back the return candidate track segments already generated in the current return cycle.

[0050] Through the above processing, the candidate segment construction module generates return candidate track segments with clear action boundaries based on the phase changes of the target wind turbine surrounding tower inspection actions, the flight mode switching before the return trigger, and the time continuity status of the evidence units in the historical navigation evidence sequence to be reviewed. The return candidate track segments carry navigation measurement result reference relationships, boundary source relationships, and segment status records. After entering the geometric closure review module, they are used to limit the comparison objects between the structural observation changes of the target wind turbine and the relative motion changes of the track.

[0051] The geometric closure verification module receives the return candidate track segments output by the candidate segment construction module and processes them using a closure verification method composed of the observed changes in the target wind turbine structure and the relative motion changes of the track derived from the return candidate track segments. This process generates a locally self-consistent contamination identification result. The geometric closure verification module processes return candidate track segments whose segment boundaries have been determined by the candidate segment construction module. It also uses the segment start boundary, segment end boundary, historical navigation evidence unit, navigation measurement result reference mark, and segment status record in the return candidate track segments.

[0052] During offshore wind turbine tower inspections, when unmanned helicopters perform blade leading edge inspections, blade root area inspections, or circumferential flight around the target wind turbine, the camera payload continuously points at the target wind turbine structure. When the unmanned helicopter is in a state of strong sea surface reflection, tower obstruction, nacelle metal reflection, or link protection flight state, the navigation measurement results can show continuous changes within local segments. However, this continuous change and the projection change of the target wind turbine structure in the payload observation record may not form a closed relationship. The geometric closure verification module reversely verifies the relative motion change of the trajectory through the observed changes of the target wind turbine structure. This is used to identify contaminated segments in which the navigation measurement results are locally self-consistent within the segment but not closed relative to the target wind turbine structure.

[0053] The geometric closure verification module first reads the segment start boundary, segment end boundary, navigation measurement result reference marker, and segment status record of the return candidate track segment. Based on the navigation measurement result reference marker, it calls the navigation measurement results corresponding to the historical navigation evidence units within the segment. The unaccepted evidence units within the segment are retained as the source of the segment status record, but are not used as the basis for continuous deduction of the relative motion changes of the track. When the return candidate track segment has a low coverage state, acceptance limit state, or missing action phase state, the corresponding state enters the subsequent closure verification admission judgment.

[0054] The geometric closure verification module reads the prior records of the target wind turbine structure and the load observation records that are in the same online recording period as the candidate return track segment. The online recording period boundary already formed by the candidate segment construction module is used to limit the time pairing relationship between the load observation records and the candidate return track segment. The prior records of the target wind turbine structure are used to limit the target wind turbine stable structural objects and the target wind turbine reference position. The target wind turbine stable structural objects include at least one of the tower centerline, nacelle outline and aviation obstruction light spatial position. When the blade root region and the outer edge of the impeller are in a locked state or a low-speed trackable state as confirmed by the current inspection task configuration, they participate in the closure verification as target wind turbine stable structural objects.

[0055] The geometric closure verification module locks the starting and ending observation frames of a segment from the load observation records based on the segment start boundary, segment end boundary, and load observation acceptance deviation. The load observation acceptance deviation is measured in seconds and is used to limit the time pairing range between the load observation frame and the segment boundary. Its source is the load observation record frame period and the online time synchronization calibration record, and it does not exceed the sum of one load observation record frame period and the upper limit of the online time synchronization error. If the segment start boundary or segment end boundary cannot be matched with a load observation frame within the load observation acceptance deviation, the geometric closure verification module marks the corresponding return candidate track segment as a closure verification restricted segment and writes the source of the missing observation frame into the local self-consistent contamination identification result.

[0056] When both the starting and ending observation frames of a segment exist, the geometric closure verification module generates structural object identifiers and projection position records in the starting and ending observation frames of the segment according to the structural object identifiers in the prior records of the target wind turbine structure. Based on the structural object identifiers, it establishes a correspondence between the same stable structural objects of the target wind turbine between the two observation frames. If the structural object identifiers cannot be uniquely matched, the stable structural objects of the target wind turbine are occluded, resulting in the inability to form projection position records, or if the correspondence between the structural objects in the two observation frames conflicts with each other, the geometric closure verification module marks the corresponding return candidate track segment as a closure verification restricted segment and does not calculate the closure deviation.

[0057] Projection position records are used to characterize the observation azimuth relationship, projection scale relationship, and structural relative order relationship of the target wind turbine stable structure relative to the load observation reference. The geometric closure verification module performs a before-and-after comparison of the projection relationship in the segment's initial observation frame and the segment's final observation frame to form the observed changes in the target wind turbine structure. In the blade leading edge inspection scenario, the changes in the projection relationship between the blade root region, the tower edge, and the nacelle outline are used as the main structural observation basis. In the tower circumferential flight scenario, the changes in the azimuth relationship between the tower edge, the nacelle outline, and the projection of the aviation obstruction lights are used as the main structural observation basis.

[0058] Using the target wind turbine reference position in the prior record of the target wind turbine structure as a relative reference object, the navigation measurement results are converted into relative position to form the relative motion change of the track. Specifically, the geometric closure verification module reads the navigation measurement results between the segment start boundary and the segment end boundary, and converts the navigation measurement results into relative azimuth change, relative distance change and circumferential direction change relative to the target wind turbine reference position. The relative motion change of the track is derived from the historical navigation evidence units within the segment that meet the time continuity requirements. Unconnected evidence units within the segment do not participate in the continuous derivation.

[0059] Before the target wind turbine structure observation changes and track relative motion changes enter the closed comparison, the geometric closure verification module performs closure verification admission judgment on the return candidate track segments based on the segment status record, the correspondence of the target wind turbine stable structure object, the load observation benchmark, and the mission navigation coordinate benchmark. If the segment status record indicates that there is a low coverage state, a limited acceptance state, or a missing action phase state, or if the correspondence of the target wind turbine stable structure object is not established, the return candidate track segment is marked as a closed verification restricted segment. If the target wind turbine reference position and the navigation measurement result are not in the same mission navigation coordinate benchmark, coordinate transformation is performed according to the coordinate mapping relationship in the current inspection mission configuration; if the coordinate transformation fails, the coordinate benchmark restricted state is written, and the return candidate track segment is marked as a closed verification restricted segment.

[0060] The load observation record also includes load line-of-sight records, gimbal attitude records, and imaging parameter records corresponding to the inspection images or video frames. Before forming the target wind turbine structure observation changes, the geometric closure verification module checks whether the load observation references corresponding to the segment start boundary and the segment end boundary are consistent. When the change in load focal length, gimbal attitude, or load line-of-sight record causes the change in projection relationship to be attributable to the motion change of the unmanned helicopter relative to the target wind turbine, the observation reference is written as restricted, and the return candidate track segment is marked as a closed verification restricted segment. During the blade leading edge inspection, when the camera load performs gimbal compensation to keep the blade edge in the center of the image, the change in the projection of the target wind turbine structure in the image can be attributed to the gimbal compensation; this situation is treated as an observation reference restricted situation to prevent the load observation reference change from being used as evidence of navigation contamination.

[0061] When the closure verification passes, a closure comparison is performed on the observed changes in the target wind turbine structure and the relative motion changes in the trajectory. The closure comparison includes directional closure comparison and radial closure comparison, which are used to compare whether the azimuth change direction of the target wind turbine stable structure in the load observation record is consistent with the direction around the tower derived from the navigation measurement results, and whether the change in the projected scale of the target wind turbine stable structure is consistent with the change in the relative distance derived from the navigation measurement results. If the navigation measurement results show that the unmanned helicopter moves along the circumference of the target wind turbine, but the target wind turbine stable structure in the load observation record does not show a corresponding change in the projected relationship, a directional closure deviation source label is written. If the navigation measurement results show that the unmanned helicopter moves away from the target wind turbine, but the target wind turbine stable structure does not show a corresponding change in the projected scale, the geometric closure verification module writes a radial closure deviation source label. The directional closure deviation source label and the radial closure deviation source label are output along with the local self-consistent pollution identification results, and are used by the return permission output module to distinguish the pollution source in the reverse use qualification determination.

[0062] To ensure the closure verification has a calculable basis, the geometric closure verification module generates closure deviations. Before generating the closure deviations, the geometric closure verification module generates... and , where subscript Indicates the observation side of the target wind turbine structure, subscript Indicates the navigation track side, The results of the circumferential orientation in the observation changes of the target wind turbine structure formed in the previous text are formed by the changes in the projected orientation of the same target wind turbine stable structure object in the first observation frame and the last observation frame of the segment. The direction of the orbit around the tower in the relative motion change of the track formed above is formed by the change in orientation of the navigation measurement results between the segment start boundary and the segment end boundary relative to the target wind turbine reference position.

[0063] The geometric closure verification module, based on the effective change boundaries of the structural side direction and the effective change boundaries of the track side direction, respectively, will... and The quantization is categorized as positive tower-circling change, no effective tower-circling change, or reverse tower-circling change, with corresponding values ​​of 1, 0, or -1. Before structural-side direction quantization, the geometric closure verification module determines the structural-side direction reference based on the mapping relationship between the load observation reference and the tower-circling direction reference in the current inspection task configuration. If the mapping relationship cannot be formed, the observation reference is written as restricted, and the return candidate track segment is marked as a closed verification restricted segment. Both positive and reverse tower-circling changes are judged based on the tower-circling direction reference in the current inspection task configuration. The effective boundary of structural-side direction change comes from the load observation error range and the stable tower-circling inspection record of similar target wind turbines in the current inspection task configuration. The effective boundary of track-side direction change comes from the navigation measurement error range and the stable tower-circling inspection record of similar target wind turbines in the current inspection task configuration. When historical samples are insufficient, the load observation error range and navigation measurement error range in the current inspection task configuration are used to form the initial boundary.

[0064] The normalized radial variation results corresponding to the observed changes in the target wind turbine structure formed earlier are generated by the changes in the projected scale or projected position of the stable structural object of the target wind turbine; within the same return candidate track segment... Using the same type of structural lateral radial observation source, when both projection scale and projection position variations are available, the structural observation priority in the current inspection task configuration is selected. Corresponding to the normalized radial variation result in the relative motion change of the trajectory formed above, which is formed by the distance change of the navigation measurement result relative to the target wind turbine reference position, the geometric closure verification module will, based on the structural side normalized amplitude limit, Convert to a dimensionless result with values ​​ranging from 0 to 1, and then, based on the track-side normalized limit, convert... The result is converted to a dimensionless result with a value range of 0 to 1. The structural normalized amplitude limiting basis is derived from the load observation error range, the observable projection scale range of the target wind turbine stable structural object, and the stable tower inspection records of similar target wind turbines in the current inspection task configuration. The track normalized amplitude limiting basis is derived from the navigation measurement error range, the target wind turbine safety distance constraint, and the stable tower inspection records of similar target wind turbines in the current inspection task configuration. When the structural normalized amplitude limiting basis or the track normalized amplitude limiting basis is missing, is 0, or cannot be formed, the geometric closure verification module does not calculate the closure deviation and marks the corresponding return candidate track segment as a closure verification restricted segment.

[0065] like and All have been formed, and the closure deviation is formed in the following manner: ; in, The deviation from closure is dimensionless and ranges from 0 to 1. The results of the observation of changes in the target wind turbine structure corresponding to the direction around the tower. The result of the trajectory around the tower is derived from the candidate trajectory segment for the return flight. Both values ​​are -1, 0 or 1. This is the directional deviation term, with a value range of 0 to 1; The normalized radial variation results corresponding to the observed structural changes of the target wind turbine. The normalized radial variation results are derived from the candidate return trajectory segments, and both values ​​range from 0 to 1. This is the radial closure deviation term, with a value range of 0 to 1; and The values ​​are dimensionless weights, all of which are greater than 0 and less than 1, and the sum of the two is 1.

[0066] The design logic of this formula is as follows: First, the directional difference between the observation side and the navigation track side of the target wind turbine structure is normalized to 0 to 1. Then, the radial difference between the two is limited to 0 to 1. The common constraint relationship between directional closure and radial closure is preserved through product suppression. Finally, the closure deviation is obtained by subtracting the degree of common closure from 1. The closure deviation is entered into the generation process of local self-consistent pollution identification results to distinguish between closure consistent segments and local self-consistent pollution segments. It also provides pollution segment basis for the reverse use qualification judgment of the return permission output module. Missing structural object correspondence, limited observation benchmark, or limited coordinate benchmark are not directly included in the calculation of closure deviation as pollution items, but are written into the local self-consistent pollution identification results as the basis for closure verification limitation.

[0067] The system reads the closure deviation and the closure deviation judgment boundary, and generates a local self-consistent contamination identification result based on the closure deviation, navigation measurement continuity, and closure verification restricted segment marking. The closure deviation judgment boundary is formed by the stable tower-circling inspection records of the same type of unmanned helicopter in the same type of target wind turbine inspection mission. Its dimensionless property takes a value greater than 0 and less than 1, and it is in the same comparison caliber as the closure deviation. When historical samples are insufficient, the initial judgment boundary in the current inspection mission configuration is adopted. The initial judgment boundary is formed by the load observation error range, navigation measurement error range, and target wind turbine safety distance constraint in the current inspection mission configuration.

[0068] When the closure deviation is not greater than the closure deviation judgment boundary, and the return candidate track segment is not marked as a closure verification restricted segment, the geometric closure verification module marks the return candidate track segment as a closure consistent segment. When the closure deviation is greater than the closure deviation judgment boundary, the navigation measurement continuity is confirmed by the time continuity status of the historical navigation evidence unit and the navigation measurement result reference mark, and the return candidate track segment is not marked as a closure verification restricted segment, the geometric closure verification module further checks whether all continuous historical navigation evidence units have valid navigation measurement result reference marks, and whether the segment status record is not written to the continuity restriction status. When the check passes, the geometric closure verification module marks the return candidate track segment as a locally self-consistent contaminated segment. When the return candidate track segment has been marked as a closure verification restricted segment, or when the closure deviation has not been formed, the geometric closure verification module marks the return candidate track segment as a closure verification restricted segment and writes it into the locally self-consistent contamination identification result, without outputting the return candidate track segment as a closure consistent segment or a locally self-consistent contaminated segment alone.

[0069] When identifying locally self-consistent contaminated segments, the segment start boundary and segment end boundary of the locally self-consistent contaminated segment are written into the contaminated boundary candidate results. If multiple consecutive return candidate track segments are marked as locally self-consistent contaminated segments, the geometric closure verification module forms the contaminated boundary candidate range with the first segment start boundary and the last segment end boundary of the consecutive contaminated segments. The closed verification restricted segments are retained as the limiting information of the contaminated boundary candidate range and are not used as positive evidence to confirm the contaminated boundary alone.

[0070] Taking the inspection of the leading edge of the target wind turbine blade as an example, when the unmanned helicopter triggers the link protection near the leeward side of the tower, the navigation measurement results show a continuous change along the circumference of the target wind turbine. However, the projection of the tower edge and the nacelle outline in the load observation record does not show a corresponding azimuth change between adjacent segments. At this time, the geometric closure verification module generates a closure deviation based on the difference between the results of the circumferential direction formed by the structural changes of the target wind turbine and the results of the circumferential direction of the flight track. When the closure deviation is greater than the closure deviation judgment boundary and the candidate flight track segment for returning is not marked as a restricted segment for closure verification, the geometric closure verification module marks the candidate flight track segment for returning as a locally self-consistent contaminated segment and writes the segment start boundary and segment end boundary into the contaminated boundary candidate result. After the contaminated boundary candidate result enters the return permission output module, it is used to limit the historical flight track segments with continuous values ​​but non-closed structures as the basis for the return calculation.

[0071] The system outputs the local self-consistent contamination identification results and sends them to the return-to-base permission output module. The return-to-base permission output module performs reverse use qualification determination on the candidate return-to-base track segments based on the local self-consistent contamination identification results, and generates return-to-base anchor points and return-to-base path generation control records.

[0072] The weights, closure deviation judgment boundaries, effective direction change boundaries, and normalized amplitude limiting bases share the same calibration basis. This calibration basis includes stable tower-circling inspection records from inspection missions involving the same type of unmanned helicopter and similar target wind turbines, the load observation error range, navigation measurement error range, and target wind turbine safety distance constraints in the current inspection mission configuration. and The contribution ratio of directional closure comparison and radial closure comparison to closure deviation is used to control the closure deviation judgment boundary, which is used to control the classification of closure consistent segments and locally self-consistent contaminated segments. The effective directional change boundary is used to exclude observation jitter and navigation noise in the directional results. The normalized amplitude limiting basis is used to form dimensionless radial change results on the structural side and the track side.

[0073] The stable tower-circling inspection record is the inspection record that has formed a closed and consistent segment and has not been written into the closed verification restricted state. Local self-consistent contamination segments and closed verification restricted segments are not included in the calibration basis. When historical samples are insufficient, the geometric closure verification module uses the initial weight configuration, initial judgment boundary, initial effective direction change boundary and initial normalized amplitude limit basis confirmed by the current inspection task configuration as the initial source. The subsequent stable tower-circling inspection records are updated after accumulation. The updated weight, closed deviation judgment boundary, effective direction change boundary and normalized amplitude limit basis are used for subsequent return cycles and do not write back the local self-consistent contamination identification results already generated in the current return cycle.

[0074] Through the above processing, the geometric closure verification module transforms the candidate return track segments generated by the candidate segment construction module into executable closure verification objects, forming the observed changes in the target wind turbine structure and the relative motion changes of the track, respectively. It also identifies locally self-consistent contamination segments that are locally continuous but not closed relative to the target wind turbine structure through the closure deviation. The local self-consistent contamination identification results are entered into the return permission output module for subsequent reverse use qualification determination and the formation of return path generation control records.

[0075] The return-to-home permission output module receives the local self-consistent contamination identification result output by the geometric closure verification module and reads the return-to-home candidate track segments output by the candidate segment construction module. Based on the local self-consistent contamination identification result, it performs reverse use qualification determination on the return-to-home candidate track segments, generates a return-to-home start-up permission result, and forms a return-to-home path generation control record based on the return-to-home start-up permission result. The return-to-home path generation control record includes the return-to-home acceptance anchor point, contamination boundary segments, and prohibited reverse use segments. It is used to limit the historical track boundaries that can be called in the return-to-home path generation process, the range of contamination boundaries that cannot be crossed, and the historical track segments that cannot be reversed. The current return-to-home cycle is a return-to-home decision processing process corresponding to the current return-to-home trigger anchor point.

[0076] In offshore wind farm inspection scenarios, when unmanned helicopters perform blade leading edge inspections, blade root area inspections, or circumferential flight around the target wind turbine, the historical track segments before the return trigger anchor point cannot all serve as reliable starting points for the return path generation. When the unmanned helicopter is on the leeward side of the tower, near the nacelle metal structure, or in the area of ​​sea surface reflection interference, the navigation measurement results can remain continuous for a local time period. However, this continuous track does not close with the observed changes in the target wind turbine structure. The return permission output module does not re-execute the geometric closure verification. Instead, it converts the closed consistent segments, locally self-consistent contamination segments, closed verification restricted segments, and contamination boundary candidate results already formed by the geometric closure verification module into control fields in the return path generation control record.

[0077] First, using the segment start boundary and segment end boundary of the return candidate track segment as an index, a correspondence is established between the local self-consistent contamination identification results and the return candidate track segments. The input sources for this correspondence are the return candidate track segments output by the candidate segment construction module, and the closed consistent segments, locally self-consistent contamination segments, closed verification restricted segments, and contamination boundary candidate results output by the geometric closure verification module. The return permission output module assigns the locally self-consistent contamination identification results under the same time boundary to the corresponding return candidate track segment according to the consistency of the segment start boundary and segment end boundary. This correspondence enters the valid segment status determination process to prevent object mismatch between the return candidate track segment and the verification result. If a certain return candidate track segment does not have a corresponding result in the local self-consistent contamination identification result, the return permission output module treats the return candidate track segment as a return candidate track segment without a corresponding verification result, so that the return candidate track segment enters the segment formation process that prohibits reverse use, and does not use the return candidate track segment as the source of the return anchor point.

[0078] Taking the circumferential flight around the target wind turbine tower as an example, the starting boundary of a certain return candidate trajectory segment corresponds to the lateral observation position of the tower, and the ending boundary of the segment corresponds to the leeward observation position of the tower. The geometric closure verification module outputs a locally self-consistent contamination segment under the same segment boundary, and the return permission output module establishes a corresponding relationship with the same segment boundary, so that the locally self-consistent contamination segment belongs to the same return candidate trajectory segment, avoiding misattributing the verification result to the adjacent blade root area inspection segment.

[0079] When the same return-to-home candidate track segment is simultaneously marked as one of two or more of the following: locally self-consistent contaminated segment, closed-loop verification restricted segment, or closed-loop consistent segment, the return-to-home permission output module determines the status of the valid segment according to the priority order of locally self-consistent contaminated segment, closed-loop verification restricted segment, and closed-loop consistent segment. This priority is not a numerical weight and does not participate in the weighted calculation. Instead, it is used to determine the eligibility for reverse use when the verification results conflict. A locally self-consistent contaminated segment indicates that the risk of contamination has been identified in the navigation measurement results, which are locally continuous but not closed relative to the target wind turbine structure. A closed-loop verification restricted segment indicates that the load observation frame is missing, the structural object correspondence fails, the coordinate reference is restricted, or the observation reference is restricted, and a basis for reverse use cannot be formed. A closed-loop consistent segment indicates that the return-to-home candidate track segment has passed the closed-loop verification.

[0080] Candidate return track segments identified as locally self-consistent contamination segments enter the formation process of contamination boundary segments and prohibited reverse-use segments; candidate return track segments identified as closed-loop verification restricted segments are not used as positive evidence of contamination, but serve as the basis for reverse-loop acceptance blocking and enter the formation process of prohibited reverse-use segments; candidate return track segments identified as closed-loop consistent segments are used to form return-loop acceptance anchor points, provided they are not blocked by locally self-consistent contamination segments, closed-loop verification restricted segments, or candidate return track segments without corresponding verification results. Through the above priority processing, the same candidate return track segment will not be simultaneously used as an acceptable segment and a prohibited reverse-use segment.

[0081] The return-to-home permission output module identifies the candidate track segment whose end boundary is no later than the return-to-home trigger anchor point and whose time distance from the return-to-home trigger anchor point is closest as the starting segment for reverse scanning. It then performs a reverse scan from the starting segment towards the historical time. The selection rule for the starting segment of the reverse scan serves as the starting point limitation for determining the eligibility for reverse use. Its input sources are the return-to-home trigger anchor point and the end boundary of the candidate track segment. No fixed time threshold is introduced. If the starting segment for reverse scanning cannot be uniquely determined due to overlapping segments, missing segments, or the inability to confirm the sequential relationship between the end boundary of the segment and the return-to-home trigger anchor point, the return-to-home permission output module will include the candidate track segment with abnormal segments in the process of forming prohibited reverse use segments and generate a return-to-home start-up permission result with a restricted reverse acceptance state. No return-to-home acceptance anchor point will be formed.

[0082] The reverse scan processes candidate track segments that have already established a correspondence with the local self-consistent contamination identification results. The time boundary of the reverse scan is jointly defined by the return trigger anchor point and the segment end boundary of the candidate track segment. It does not call track records formed after the return trigger anchor point, and does not modify the reverse use qualification determination before the current return trigger anchor point based on the subsequent return execution results.

[0083] During the reverse scan, the return-to-home permission output module stops scanning when it encounters a locally self-consistent contaminated segment, a closed verification restricted segment, or a return-to-home candidate track segment without corresponding verification results. The closed and consistent segment before the stop is used to generate the return-to-home anchor point. This stopping rule is used to prevent the return-to-home path generation process from calling earlier historical tracks by crossing contaminated or verification restricted segments. If the starting segment of the reverse scan has been identified as a locally self-consistent contaminated segment, a closed verification restricted segment, or a return-to-home candidate track segment without corresponding verification results, the return-to-home permission output module does not form a return-to-home anchor point and includes the starting segment of the reverse scan in the process of forming segments that are prohibited from reverse use.

[0084] If no locally self-consistent contaminated segments, closed verification restricted segments, or return candidate track segments without corresponding verification results appear between the starting segment of the reverse scan and the starting segment of the segment boundary sequence, the return permission output module uses the return candidate track segments with the valid segment status of closed consistent segments within this range to generate the return acceptance anchor point. If there are no closed consistent segments before stopping, and there are no valid return candidate track segments with the valid segment status of closed consistent segments between the starting segment of the reverse scan and the starting segment of the segment boundary sequence, the return permission output module generates the return start permission result with restricted reverse acceptance status.

[0085] Taking the scenario of transitioning to the return preparation state after blade leading edge inspection as an example, the closest candidate return track segment before the return trigger anchor point is marked as a locally self-consistent contamination segment. The navigation measurement results within this segment show that the unmanned helicopter moves continuously along the circumference of the target wind turbine, while the tower edge and nacelle outline in the load observation record do not show corresponding projection changes. The return permission output module does not cross this locally self-consistent contamination segment and selects an earlier closed and consistent segment as the return acceptance anchor point, thereby avoiding using historical tracks with continuous values ​​but non-closed structures as the basis for return calculation.

[0086] Among the candidate return track segments that can be used to generate a return anchor point and whose effective segment status is a closed and consistent segment, the return permission output module determines the target closed and consistent segment according to the segment boundary sequence. The segment boundary sequence comes from the segment encapsulation boundary sequence formed by the candidate segment construction module and is used to limit whether the closed and consistent segments are continuously connected in time. The target closed and consistent segment is the closed and consistent segment whose segment end boundary is closest to the return trigger anchor point among the closed and consistent segments that can be used to generate a return anchor point. The return permission output module forms the return anchor point together with the segment end boundary of the target closed and consistent segment and the navigation measurement result reference mark corresponding to the historical navigation evidence unit that is located within the target closed and consistent segment, no later than the segment end boundary, and has the closest time distance to the segment end boundary and has a valid time connection status.

[0087] The segment end boundary is used to limit the historical track time boundary that the return path generation can accept. The navigation measurement result reference mark is used to limit the source of navigation measurement results that can be called in the return path generation. If there is no historical navigation evidence unit with a valid time acceptance state in the target closed and consistent segment, the return permission output module will not use the corresponding segment end boundary as a separate return acceptance anchor point, and will generate a return start permission result with a restricted reverse acceptance state.

[0088] The return-to-base permission output module forms a contaminated boundary segment based on the contaminated boundary candidate results and the segment boundaries of the locally self-consistent contaminated segments. Specifically, the return-to-base permission output module reads the contaminated boundary candidate results and uses the segment start boundary and segment end boundary of the locally self-consistent contaminated segment as the boundary source of the contaminated boundary segment. If multiple consecutive return-to-base candidate track segments are determined to be locally self-consistent contaminated segments under the segment boundary sequence relationship, a continuous contaminated boundary segment is formed by the first segment start boundary and the last segment end boundary of the continuous contaminated segment. After the contaminated boundary segment enters the return-to-base path generation control record, it is used to limit the return-to-base path generation process from crossing the locally self-consistent contaminated range. The closed-loop verification restricted segment is not used as positive evidence of contamination. When the closed-loop verification restricted segment and the contaminated boundary segment are adjacent in the segment boundary sequence, the closed-loop verification restricted segment is retained as the restricted boundary source of the contaminated boundary segment.

[0089] The following return path candidate segments that are blocked from being reversed, return path candidate segments whose boundaries overlap with the boundaries of contaminated boundary segments, and return path candidate segments that are interrupted by contaminated boundary segments, closed verification restricted segments, or no corresponding verification results when reversed from the return trigger anchor point, are written into the prohibited reverse use segment. The prohibited reverse use segment is used to block the return path generation module from calling historical track segments that have the risk of contamination, closed verification restricted, or no corresponding verification results.

[0090] The return-to-origin permit output module generates a return-to-origin permit calculation result based on the return-to-origin accepting anchor point, contaminated boundary segments, and prohibited reverse-use segments. When a return-to-origin accepting anchor point is formed, and there are no contaminated boundary segments, closed-loop verification restricted segments, or return-to-origin candidate track segments without corresponding verification results between the return-to-origin accepting anchor point and the return-to-origin triggering anchor point, the return-to-origin permit output module generates a return-to-origin permit calculation result in a restricted reverse-origin accepting state. After the return-to-origin permit calculation result is formed, prohibited reverse-use segments shall not be used as the historical path calculation basis for the return-to-origin path generation process.

[0091] Based on the return-to-origin permit result, a return-to-origin path generation control record is formed. The return-to-origin path generation control record is indexed according to the return-to-origin trigger anchor point, the segment start boundary and segment end boundary of the return-to-origin candidate track segment. It is used to constrain the scope of historical track segments called by the return-to-origin path generation module. The return-to-origin accepting anchor point is used to limit the historical track boundary that can be accepted when the return-to-origin path is generated. The contaminated boundary segment is used to limit the return-to-origin path generation process from crossing the local self-consistent contamination range. The prohibited reverse use segment is used to block the return-to-origin path generation module from calling historical track segments that have contamination risk, limited closure verification, or no corresponding verification result. The return-to-origin path generation control record is only formed corresponding to the current return-to-origin trigger anchor point and does not rewrite the closure deviation judgment boundary, direction effective change boundary and normalized amplitude limit basis in the geometric closure verification module.

[0092] If the local self-consistent contamination identification result only contains closed-loop verification restricted segments and not local self-consistent contamination segments, the return-to-home permission output module will not treat the closed-loop verification restricted segments as local self-consistent contamination segments, nor will it use the closed-loop verification restricted segments as the source of the return-to-home anchor point. When a new load observation record or a new local self-consistent contamination identification result is generated in a subsequent return-to-home cycle, the return-to-home permission output module will re-execute the reverse use qualification determination in the subsequent return-to-home cycle. The return-to-home path generation control record already generated in the current return-to-home cycle will not be reverse-written by subsequent records.

[0093] Through the above processing, the return permission output module transforms the local self-consistent contamination identification results into the return acceptance anchor point, contamination boundary segment, and prohibited reverse use segment in the return path generation control record. This allows the return path generation process to avoid local self-consistent contamination segments and restricts closed verification restricted segments and return candidate track segments without corresponding verification results as historical starting points.

[0094] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An autonomous return-to-home decision-making system for unmanned helicopters, characterized in that, include: The navigation evidence encapsulation module acquires navigation evidence records formed by the offshore wind power inspection drone helicopter before the return trigger anchor point. The return trigger anchor point corresponds to the autonomous return trigger time. The module uses the evidence encapsulation method with the return trigger anchor point as the retrospective endpoint to generate a sequence of historical navigation evidence to be reviewed. The navigation evidence records represent the correspondence between the stage of the inspection around the target wind turbine and the navigation measurement results in the same retrospective period. The candidate segment construction module uses the tower-around action boundary segmentation method to process the historical navigation evidence sequence to be reviewed and generate return candidate track segments. The tower-around action boundary is formed by the changes in the stage of the tower-around inspection action around the target wind turbine and the flight mode switching before the return is triggered. The geometric closure verification module adopts a closure verification method consisting of the relative motion changes of the trajectory derived from the observed changes in the target wind turbine structure and the candidate trajectory segments of the return flight. It processes the candidate trajectory segments of the return flight and generates local self-consistent pollution identification results. The return-to-base permit output module performs reverse use qualification determination on the candidate return-to-base track segments based on the local self-consistent pollution identification results, generates the return-to-base start-up permit result, and writes the candidate return-to-base track segments into the return-to-base path generation control record based on the return-to-base start-up permit result. The return-to-base path generation control record includes the return-to-base acceptance anchor point, pollution boundary segments, and prohibited reverse use segments.

2. The autonomous return-to-home decision-making system for unmanned helicopters according to claim 1, characterized in that, The specific operation of the evidence encapsulation method is as follows: read the airborne return trigger record used to form the return trigger anchor point, and take the record time of forming the autonomous return command as the return trigger anchor point; take the return trigger anchor point as the backtracking endpoint, and generate the backtracking period based on the base duration of the tower-circling inspection action in the current inspection task configuration and the readable record coverage duration before the return trigger; During the retrospective period, candidate records belonging to the same target wind turbine surrounding tower inspection action phase are extracted. Based on the task attribution of the candidate record relative to the return trigger anchor point and the execution evidence attribution of the action phase, the candidate records with the established attribution relationship are marked as action phase locked records, and the candidate records with the invalid attribution relationship are marked as non-locked evidence records.

3. The autonomous return-to-home decision-making system for unmanned helicopters according to claim 2, characterized in that, The process of generating the historical navigation evidence sequence to be reviewed is as follows: The execution time of the action phase lock record is encapsulated, and the action observation time is matched with the navigation measurement time in the navigation measurement record within the allowable time deviation. The allowable time deviation is determined based on the airborne recording time synchronization error and the sampling period of the navigation measurement record. When all navigation measurement times adjacent to the action observation time exceed the allowable time deviation, a low-frequency measurement marker is written, and the corresponding candidate record is encapsulated as an unencapsulated evidence unit. When a match is found, an evidence acceptance value is generated according to the time deviation level between the action observation time and the navigation measurement time, and the matched candidate record is encapsulated as a historical navigation evidence unit. The navigation evidence encapsulation module arranges the historical navigation evidence units and unencapsulated evidence units according to the chronological order of the action observation times, generating the historical navigation evidence sequence to be reviewed.

4. The autonomous return-to-home decision-making system for unmanned helicopters according to claim 3, characterized in that, The boundary candidate generation operation of the tower-around action boundary segmentation method is as follows: scan the historical navigation evidence sequence to be reviewed in chronological order of action observation time. The evidence units in the historical navigation evidence sequence to be reviewed include historical navigation evidence units and unaccepted evidence units. Based on the action observation time of the evidence unit, corresponding candidate records and flight control mode records before the return trigger anchor point are obtained within the same online recording period. The corresponding candidate records form the action phase type, and the flight control mode records form the flight mode field. The action phase transition and flight mode switch are used as the trigger conditions for the tower-around action boundary. The action phase transition is jointly confirmed by the action phase type of the adjacent evidence unit and the action phase continuous confirmation condition, and the flight mode switch is confirmed by the flight mode field before the return trigger anchor point. When the tower-around action boundary trigger condition is formed, the action observation time corresponding to the trigger condition is written into the tower-around action boundary candidate record. The action phase continuous confirmation condition is formed by the action record sampling period and the shortest action hold time in the current inspection task configuration.

5. The autonomous return-to-home decision-making system for an unmanned helicopter according to claim 4, characterized in that, The process for generating candidate return trajectory segments is as follows: Based on the temporal continuity status of evidence units in the historical navigation evidence sequence to be reviewed, boundary restriction processing is performed on candidate records of the tower-circling action boundary to form internal boundary records. The temporal continuity status is jointly characterized by the evidence continuity value of the historical navigation evidence unit and the non-continuation status of the non-continuation evidence unit. The internal boundary records within the same action record sampling period are merged and combined with the sequence start boundary and sequence end boundary of the historical navigation evidence sequence to be reviewed to form a segment encapsulation boundary sequence. The sequence start boundary and sequence end boundary are formed by the action observation times of the first and last evidence units in the historical navigation evidence sequence to be reviewed, respectively. Fragment encapsulation is performed on the evidence unit set between adjacent fragment encapsulation boundaries. When the evidence unit set contains only unaccepted evidence units, a fragment insufficient acceptance record is written. When the evidence unit set contains historical navigation evidence units, the fragment is encapsulated to generate a return candidate track fragment. Unaccepted evidence units are not considered as valid continuous evidence of relative motion changes within the track.

6. The autonomous return-to-home decision-making system for an unmanned helicopter according to claim 5, characterized in that, The operations for forming the observed changes in the target wind turbine structure and the relative motion changes in the track are as follows: Read the segment start boundary, segment end boundary, navigation measurement result reference marker, and segment status record of the return candidate track segment, and call the navigation measurement results corresponding to the historical navigation evidence unit within the segment based on the navigation measurement result reference marker; read the prior record of the target wind turbine structure and the load observation record in the same online recording period as the return candidate track segment, perform a projection relationship comparison on the target wind turbine stable structure object corresponding to the segment start boundary and segment end boundary to form the observed changes in the target wind turbine structure; using the target wind turbine reference position in the prior record of the target wind turbine structure as the relative reference object, perform relative position conversion on the navigation measurement results to form the relative motion changes in the track.

7. The autonomous return-to-home decision-making system for an unmanned helicopter according to claim 6, characterized in that, The specific operation of the closed-loop verification method is as follows: based on the segment status record, the correspondence between the target wind turbine stable structure object and the mission navigation coordinate reference, the closed-loop verification admission judgment is performed on the return candidate track segment; when the closed-loop verification admission judgment fails, the return candidate track segment is marked as a closed-loop verification restricted segment; When the closure verification and access judgment is passed, directional closure comparison and radial closure comparison are performed on the observed changes in the target wind turbine structure and the relative motion changes of the trajectory to form the closure deviation.

8. The autonomous return-to-home decision-making system for unmanned helicopters according to claim 7, characterized in that, The operation to generate the local self-consistent pollution identification result is as follows: read the closure deviation amount and the closure deviation judgment boundary. The closure deviation judgment boundary is formed by the stable tower inspection record of the same type of unmanned helicopter in the same type of target wind turbine inspection mission. When the historical samples are insufficient, the initial judgment boundary in the current inspection mission configuration is used. When the closure deviation is not greater than the closure deviation judgment boundary and the return candidate track segment is not marked as a closure verification restricted segment, the return candidate track segment is marked as a closure consistent segment; When the closure deviation is greater than the closure deviation judgment boundary, the continuity of navigation measurement is confirmed by the time-carrying state and the navigation measurement result reference mark, and the return candidate track segment is not marked as a closure verification restricted segment, the return candidate track segment is marked as a locally self-consistent contaminated segment, and the segment start boundary and segment end boundary of the locally self-consistent contaminated segment are written into the contaminated boundary candidate result.

9. The autonomous return-to-home decision-making system for an unmanned helicopter according to claim 8, characterized in that, The specific operation for reverse use qualification determination is as follows: using the segment start boundary and segment end boundary of the return candidate track segment as indexes, a correspondence is established between the local self-consistent contamination identification result and the return candidate track segment; when the same return candidate track segment is simultaneously marked as two or more of the following: local self-consistent contamination segment, closed verification restricted segment, and closed consistency segment, the valid segment status is determined according to the priority order of local self-consistent contamination segment, closed verification restricted segment, and closed consistency segment; the return candidate track segment whose segment end boundary is no later than the return trigger anchor point and whose time distance from the return trigger anchor point is closest is determined as the reverse scan start segment, and a reverse scan is performed from the reverse scan start segment towards the historical time direction; the scan stops when the first local self-consistent contamination segment or closed verification restricted segment appears, and the closed consistency segment before the stop is used to generate the return acceptance anchor point.

10. The autonomous return-to-home decision-making system for an unmanned helicopter according to claim 9, characterized in that, The generation operations for the return-to-base start-up permission result and the return-to-base path generation control record are as follows: In the closed and consistent segment before stopping, the target closed and consistent segment is determined according to the segment boundary sequence. The return-to-base anchor point is formed by the segment end boundary of the target closed and consistent segment and the navigation measurement result reference mark corresponding to the historical navigation evidence unit with valid time continuity adjacent to the segment end boundary. The contaminated boundary segment is formed based on the contaminated boundary candidate result and the segment boundary of the locally self-consistent contaminated segment. The prohibited reverse-use segment is formed based on the return-to-base candidate track segment corresponding to the stop of the reverse scan and the return-to-base candidate track segment that is interrupted by any segment of the contaminated boundary segment or the closed verification restricted segment of the reverse call path. The return-to-base permission output module generates the return-to-base start-up permission result based on the return-to-base anchor point, the contaminated boundary segment, and the prohibited reverse-use segment, and writes the return-to-base anchor point, the contaminated boundary segment, and the prohibited reverse-use segment into the return-to-base path generation control record.

Images